Here's the 5 times leverage notebook:

Here's a brief analysis:

From baseline annual returns of 2.9% to 13.2% at 5x leverage, a 4.55 times increase (not quite 5x)

From baseline annual volatility of 1.6% to 6.9% at 5x leverage, a 4.3125 times increase

From baseline maximum drawdown of -0.7% to -3.6% at 5x leverage, a 5.14 times increase

Sharpe Ratio is about the same, from 1.8 to 1.84

The rest of the metrics and its 5x leverage impact can be seen in the tearsheet.

So the main takeaway here is the emphasis on low volatility which leads me to question the validity of flooring the 63 day rolling volatility at 2% in the contest scoring. I have asked Q this question before as to what their basis for flooring volatility at 2% and their response was that it was a guard for those who are trying to game volatility to get a high score and the basis of that number, 2% was based on theirstudy and community feedback. That's fair enough but now that I am looking at the results of my algo which has volatility of less than 2% at 1 unit leverage, I realized that Q might be missing out on a good thing given that their fund execution of multiple leveraging highly relies in low volatility.

James,

I totally agree with you regarding the volatility floor. Based on Jess' recent algo feedback webinar, it seems that the real focus for the Fund is super-diversified, low position concentration algos with very low volatility. I think the floor on volatility is actually counterproductive in identifying these types of algorithms through the contest, especially during what is a (relatively) very low volatility market regime.

Because of this, I was incentivised to enter a version of my algorithm into the contest which has greater position concentration, and lo and behold, it is performing "better" than my less concentrated strategy.

I would be interested to hear from others whether they have changed their approach in light of the new contest, and whether they feel the new contest accurately reflects the type of strategies that Quantopian are seeking for their Fund.

Will

Hi Will,

The volatility floor of 2% distorts scoring of truly low volatility strategies that churn out decent, legitimate returns at very low position concentration. It is counterproductive specially if their main reason for flooring volatility is to guard against people "gaming" volatility (i.e. using only small portion of investible capital to trade and parking the rest in cash). I'm sure there are other or better ways to guard against gaming.

Just to illustrate:

Algo 1 score = returns @2.5% / volatility @ 2.5% = 1.0
Algo 2 score = returns @2% / volatility @1.9% = 1.0 (because of 2% floor and score should be 1.053 without floor)

Now, with algo execution at several times leverage, I would (should) choose Algo 2!

I am actually withholding entering my version in the contest because its volatility is below 2% and would not be optimally scored. What a waste of what could be a good thing.

Hi Will and James,

I agree as well. I’m effectively incentivised to enter a more concentrated strategy so as to not hit the imposed 2% volatility floor.

@Jamie McCorriston,

I think I understand the reasoning for the floor (spoofing low volatility during the backtests) but perhaps it can be removed for algos running live in the contest for 63 days or longer, as they are much more likely to have ‘real’ 63 day rolling volatility? Otherwise you’re pretty much guaranteed to have a floor on position concentration as well, no?